Do Public Works Programs improve Child Education: Evidence from India Sonalde Desai, Omkar Joshi and Reeve Vanneman Introduction and Background Public employment programmes are often used as policy instruments for poverty reduction. However, such programmes also have a multiple ripple effect in that they not just affect poverty but also influence other socio-economic variables like women empowerment and educational performance of the children positively. Mahatma Gandhi National Rural Employment Guarantee Act (MGNREGA), passed in 2005, is one of the largest public works programme in India. MGNREGA promises 100 days of guaranteed wage employment in a financial year to every rural household whose adult members volunteer to do unskilled manual work. The act initially covered only 200 rural districts, however was extended subsequently to additional 130 districts in 2007-08 and with effect from 1st April 2008 was implemented in the entire country. MGNREGA is first of its kind and offers a marked paradigm shift from previously launched wage employment programmes in India, i that it pro ises the right to ork as a legal right. The act is unique in its design which is bottom-up, people-centric, demand driven, self-selecting and rights-based design. Two critical objectives of the act are: 1. Ensuring livelihood security for the most vulnerable people, those living rural areas 1, through providing employment opportunities for unskilled manual work; and 2. Aiding in the empowerment of marginalized communities, especially women, Scheduled Castes (SCs) and Scheduled Tribes (STs), through the processes of rights- based legislation. Although improvement in child well-being is not an objective of the MGNREGA, it is precisely for that reason that MGNREGA is a useful marker in understanding the relationship between income security a d hildre s edu atio al out o es. in a rapidly growing economy like India, educational performance of young children is critical for truly realizing the potential of the demographic dividend and for enrichment of human capital, which is crucial for sustained economic growth. The last decade saw increase in various government initiatives in the field of child education. Sarva Shiksha Abhiyan (SSA) was started as a mission-mode programme for universalization of elementary education. Right to Education was passed and was made a Fundamental Right. As a result of these initiatives, India has achieved near-universal rates of school enrolment in the last decade. Despite the high rates of school enrolment, repeated rounds of Annual Status of Education Report (ASER) surveys have highlighted that the learning levels of Indian children have continued to remain low and in fact have declined marginally. 2 1 This focus on vulnerable population was enhanced through phased implementation with the first 200 districts chosen based on their backwardness. 2 ASER 2015
MGNREGA and Educational Outcomes: Employment guarantee programs affect educational outcomes in two ways. They increase incomes and in particular, ensure income smoothing throughout the year. But at the same time, availability of these opportunities also imposes a time cost. Employment of women in MGNREGA has been rising sharply and today more than 50% of the participants are women. This may increase demands on children for domestic chores as well as childcare for younger siblings. Thus, whether MGNREGA is net e efit for hildre s edu ation remains an open question. In the literature, unlike some of the other areas of impact of MGNREGA like poverty, wages and employment, effe t of MGNE GA o hildre s well-being and educational performance has received less attention. Uppal (2009), using the Young Lives Data of Andhra Pradesh examines whether MGNREGA is acting as a safety net for children. The study finds a positive correlation between MGNREGA and anthropometric measures of children as indicators of health. Afridi et. al (2012) using the sa e You g Li es pa el data look at i pa t of MGN EGA o hildre s edu atio al attai e t ia o e s i reased a ess to la our arket opportu ities. The o e up ith a i fere e that a rise i the other s share of pare tal MGNREGA workdays increases the school attendance and grade attainment of her children, particularly girls. Mani et. al (2014) using the quasi-experimental framework and the same Young Lives panel find a strong positive effect of MGNREGA participation on grade progression, reading and math scores. While all these studies make an important contribution, they make use of the Young Lives panel, which is located in a single district in a single state. Our paper contributes to this existing literature by studying the relationship betwee MGN EGA parti ipatio a d hildre s edu atio al perfor a e at the national level thereby capturing a wide variation in the implementation of MGNREGA across differe t states i I dia. It also akes i porta t o tri utio dise ta gli g the i di idual i o e effe t e a ati g fro the household parti ipatio i the progra e fro the o ial effe t that arises from overall changes at the village level. Research Question This paper examines the impact of MGNREGA participation by a household on the educational perfor a e of hildre. pe ifi all, e look at the follo i g i di ators of hildre s edu atio al performance- a Childre s o pleted ears of s hooli g Childre s edu atio al e pe diture Childre s ti e i s hool related a ti ities d Childre s a ilit to read a d do ath. We also e a i e the differe t path a s through hi h MGN EGA ould affe t hildre s edu atio al perfor a e. Data The data in this paper come from the nationally representative multi- topic India Human Development Survey (IHDS). This panel survey was conducted in two rounds, the first in 2004-05 and the latest in 2011-12. IHDS-I and IHDS-II are part of a collaborative research programme between the National Council of Applied Economic Research (NCAER) and the University of Maryland with the goal to document changes in the daily lives of Indian households in the face of a society undergoing rapid transition. IHDS was conducted in all states and union territories in India except the union territories of Andaman and Nicobar Islands and Lakshadweep. It has gathered detailed village, household and individual information about a range of demographic and socio-economic variables viz. income, employment, consumption expenditure, education, gender relations, social network, marriage, youth, health and fertility.
This paper uses rural household data from both the rounds of the survey. Methodological Challenges to Impact Evaluation: Assessing the impact of any program is a difficult task, since we can never know what the situation would have been in absence of the program. For example, if NREGA pays Rs. 130 Per day to the workers, it does not mean that income of the individuals went up by 130 Rs. If they are diverted from other lower paying work (e.g. as a manual labourer earning Rs. 75 per day), the increase in their income is only Rs. 55. Assessing the impact of NREGA on household well-being is even more complicated. Since the programme offers manual work, it is typically used by individuals unable to find higher paying employment, making it difficult for to evaluate its impact. For example, NREGA may be of particular importance to adivasis who live in districts like Mandla or Dang with few income opportunities. Even if MGNREGA improves their opportunities, external circumstances may still not allow to them to catch up with residents of better-off districts like Jabalpur or Vadodara. In that case, we really need to compare any improvement in their lives in relative terms. We would not expect the lives of adivasis to be better than those of forward castes due to MGNREGA; we need to examine whether access to MGNREGA has improved their welfare from what it would have been without the programme. This requires a comparison of households before and after the implementation of the programme and comparison of this improvement, or lack thereof, between households with and without MGNREGA participation. This method, known as difference-in-difference method is extensively used in impact evaluation 1. We anticipate two types of effects of MGNREGA: Individual Effect: Household incomes may increase due to the implementation of MGNREGA. However, above and beyond pure income increase, the demand driven nature of MGNREGA provides work to households during periods of low agricultural demand. This would allow households to smooth consumption throughout the year and provide income during emergencies like droughts and floods as well as temporary or permanent unemployment. Social Effect: Fortunes of families in a village are often tied together. In villages where destitution prevails, few banks will set up branches, allowing moneylenders to rule. Growth of MGNREGA may encourage creation of local branches and weaken the hold of moneylenders, benefitting both MGNREGA participants and non-participant. If the social audit process encourages honesty and commitment among gram panchayat leaders, it will increase accountability not only in MGNREGA program but also among government school teachers and doctors. We also note that, presence of MGNREGA work is associated with a modest increase in private sector wages, which benefits both MGNREGA participants and non-participants. Thus, we may find that having MGNREGA programme implemented in a village may affect the lives of both participants and non-participants through transforming the social and economic fabric of the village society. We may miss out capturing this social effect if we only compare participants and non-participants.
We address these methodological challenges by dividing our sample in three categories, exposed to three different levels of MGNREGA program intensity: 3 1. Households living in low intensity villages: We define those villages in which no member of the IHDS sample participated in MGNREGA as being low intensity villages. Since about one in four rural households participate in MGNREGA, we would normally expect about four to five households to be working for MGNREGA in our IHDS sample of about 20 households per village. If no household is participating, it reflects either poor demand for MGNREGA (as in richer states like Gujarat) or poor administration of MGNREGA (as in states like Bihar). 4 2. Non-participant households in participant villages: These households live in villages where programme is being implemented but the index household did not participation in the prior year. Comparison between low intensity villages and non-participant households in participant villages allows for an estimate of the social effect. 3. Participating household: This group consists of households that participated in MGNREGA in the year preceding the survey. The difference between participating households and non-participating households in participant villages provides estimate of individual effect, while the difference between these households and those living in low intensity villages provides an estimate of total effect. Since some households in low intensity villages may still be doing MGNREGA work (and hence may benefit from the social effect), this is a highly conservative estimate of the total effect. Results We observe that with MGNREGA participation all the indicators of educational performance of children improved for MGNREGA participating households. Children from MGNREGA households are more likely to attain higher education and have improved learning outcomes than their peers from the non-mgnrega households. 3 In each case, although we present basic descriptive statistics for simplicity, significance test for the difference-in-difference (i.e. the interaction term) is conducted while controlling for income, village development level, social group and other relevant variables in linear probability models. 4 It is possible that households outside our sample may participate in MGNREGA and there may indeed be some MGNREGA activity in what we define as low intensity villages. But if so, observed differences between these villages and participant villages would be even greater than we observe if we could limit our comparison group to villages with no MGNREGA activity.
Table 1: Cha ges i Childre s Edu atio al perfor a e a o g MGNREGA parti ipa ts y village level of MGNREGA Participation. Outcome Variable 2004 05 2011 12 Difference Standards completed (ages 6 14) Low MGNREGA participation village 3.43 3.87 0.44 Difference-indifferences Significance for D-I-D village 3.14 3.59 0.45 0.01 * MGNREGA participant households 3.00 3.74 0.74 0.30 *** Can read a paragraph (ages 8 11) Low MGNREGA participation village 55.6 49.0-6.58 village 50.7 49.4-1.34 5.24 ** MGNREGA participant households 40.3 43.1 2.80 9.38 *** Can subtract two digit numbers (ages 8 11) Low MGNREGA participation village 48.2 43.3-4.84 village 43.8 40.6-3.18 1.66 MGNREGA participant households 34.6 36.0 1.43 6.27 *** Educational expenses (ages 6 14) Low MGNREGA participation village 1393 2411 1018 village 1428 2212 784-234 ** MGNREGA participant households 911 1377 466-551 *** Participate in wage work (ages 11 14) Low MGNREGA participation village 2.1 1.9-0.252
village 3.0 2.1-0.892-0.640 *** MGNREGA participant households 5.9 4.2-1.661-1.409 *** Hours spent in school, doing homework and at tuition (ages 6 14) Low MGNREGA participation village 33.5 37.4 3.9 village 31.1 37.0 5.8 1.9 *** MGNREGA participant households 29.8 37.0 7.2 3.3 *** What accounts for these improvements in educational outcomes? One might assume that incomes from MGNREGA might be funnelled into buying books or getting private tuitions for children, thereby improving their education. But comparison of educational expenditures, enrolment in private schools and access to private tutoring seem not to benefit from NREGA participation. While financial i est e ts i hildre s edu atio ha e i reased for hildre i MGN EGA households, the ha e increased even more for non-participating families in the other two groups, both participant and nonparticipant households, this increase is far greater for non-participants, which in turn widens the gap between the three groups instead of narrowing it. Hence it would be hard to attribute improving educational outcomes to greater education investments as a result of MGNREGA. o ho do e a ou t for the e efi ial effe ts o hildre s edu atio e o ser e? The answer seems to lie in the amount of time children spend in school and school related activities. The IHDS asked questions about number of hours children spent in school, in doing homework and in attending tuition per week. Table 6.2 shows the results from this analysis. In 2004-5, children from NREGA households spent on average 4 hours less per week in educational activities compared to those in low intensity villages and one our less than their non-participating neighbours. In 2011-12, they had caught up. It maybe that NREGA helps reduce child labour, thereby improving educational outcomes. Although child labour is difficult to measure and available statistics show only a very small percent of children as participating in wage work, for those children employed in these activities it could be a substantial time burden. IHDS shows that about 6% of the children ages 11-14 were engaged in wage work in 2004-5 among NREGA households, but this proportion dropped to 4% in 2011-12, while the proportion in labour force among non-participants held steady at about 2-3%. Formal labour force participation may be the tip of the iceberg. If we think about the myriad of counted and uncounted activities poor children are expected to engage in, it is not surprising that income security for households could lead to decline in time demands on children and increase their educational outcomes. 5 5 We also e a i ed ha ges i hildre s utritio al status i the o te t of MGN EGA parti ipatio. However, although MGNREGA participation is associated with a decline in severe stunting (low height-for-age), this relationship is not statistically significant and not reported here.
We are in the process of examining different aspects of educational inputs and expect that the final paper will provide a better understanding of the pathways through which public works programs influence child education. References: Afridi Farzana, Abhiroop Mukhopadhyay, and Soham Sahoo (2012): Fe ale La our For e Participation and Child Education in India: The Effect of the National Rural Employment Guarantee he e. I Institute for the Study of Labor (IZA) Discussion Paper No. 6593, IZA, Bonn ASER Centre (2015): ASER 2014: Annual Status of Education Report. New Delhi. ASER Centre Mani, Subha, Jere R. Behrman, Shaikh Galab, and P. Prudhvikar Reddy (2014 I pa t of the N EG o hooli g a d I telle tual Hu a Capital. GCC Worki g Paper eries, GCC 14-01. Uppal, Vi a ak 2009 : Is the N EG a afet Net for Childre? tud i g A ess to the Natio al ural Employment Guarantee Scheme for Young Lives Families and Its Impact on Child Outcomes in Andhra Pradesh. Dissertatio, Master of ie e i E onomics for Development, University of Oxford.